Human lower extremity joint moment prediction: A wavelet neural network approach

Abstract

Joint moment is one of the most important factors in human gait analysis. It can be calculated using multi body dynamics but might not be straight forward. This study had two main purposes; firstly, to develop a generic multi-dimensional wavelet neural network (WNN) as a real-time surrogate model to calculate lower extremity joint moments and compare with those determined by multi body dynamics approach, secondly, to compare the calculation accuracy of WNN with feed forward artificial neural network (FFANN) as a traditional intelligent predictive structure in biomechanics.To aim these purposes, data of four patients walked with three different conditions were obtained from the literature. A total of 10 inputs including eight electromyography (EMG) signals and two ground reaction force (GRF) components were determined as the most informative inputs for the WNN based on the mutual information technique. Prediction ability of the network was tested at two different levels of inter-subject generalization. The WNN predictions were validated against outputs from multi body dynamics method in terms of normalized root mean square error (NRMSE (%)) and cross correlation coefficient ( ).Results showed that WNN can predict joint moments to a high level of accuracy (NRMSE<10%, >0.94) compared to FFANN (NRMSE<16%, >0.89).A generic WNN could also calculate joint moments much faster and easier than multi body dynamics approach based on GRFs and EMG signals which released the necessity of motion capture. It is therefore indicated that the WNN can be a surrogate model for real-time gait biomechanics evaluation.

Estimación de par basada en electromiografía de superficie: potencial herramienta para la rehabilitación de rodilla

“…De acuerdo con la literatura, se han desarrollado múltiples algoritmos que permiten la estimación del par de la rodilla utilizando señales de sEMG provenientes de los músculos asociados a la flexión y extensión de la rodilla. El 23% de los algoritmos lo utilizan para mediciones del par en condiciones estáticas (33,38,40,42), mientras que, para la estimación del par durante la ejecución de movimientos de la rodilla el porcentaje de algoritmos es del 77% (32, 34,35,[39][40][41][43][44][45][47][48][49]. De acuerdo con la revisión de la literatura, para el desarrollo de estos algoritmos se han uti- En la actualidad existen métodos, basados en análisis biomecánicos y leyes físicas de la dinámica del movimiento para el cálculo del par ejercido por un sujeto, durante movimientos de flexión y extensión de rodilla (55).…”

The Application of NCaRBS to the Trendelenburg Test and Total Hip Arthroplasty Outcome

“…Three models were considered in the NN analyses, when one, two and three hidden nodes were employed in the hidden layer1 (with three outputs in the output layer). Table 2 breaks down the actual and predicted classification results for the three cases.…”

“…To offer a level of comparison with the NCaRBS results, two techniques previously employed in studies involving biomechanical data were also considered, namely linear discriminant analysis (LDA)21,28 and neural networks (NNs) 1,3,16. Comparisons were made with respect to the level of correct classification when using LDA, and level of correct classification and fit in the case of NNs.…”

Prediction of active peak force using a multilayer perceptron

“…Thus, both the general approximation of neural networks and the localization property of wavelets were utilized for the model design. The model was reported to be stable, robust, high level of accuracy and fast [8]. In [9], force magnitude, stiffness, and elastic energy based on full body kinematics during running were estimated.…”

“…The predicted GRFs and moments were the model outputs. In [8], a three-layer neural network with wavelet activation functions in hidden layers was used. Thus, both the general approximation of neural networks and the localization property of wavelets were utilized for the model design.…”